Temperatures, Clock Rates & Overclocking
Overclocking & Undervolting
Conventional overclocking through a higher power limit and more aggressive clock rate is a dead-end. Brute force just isn't the answer. Because Asus had to follow AMD's guidelines, this implementation is already running at its limit. Sure, you could dial in higher fan speeds to cool things down, creating more noise in the process, but who really wants that? As we explained in AMD RX Vega 64: The Tom's Hardware Liquid Cooled Edition, even with higher frequencies and brutal power adjustments, it is almost impossible to get Radeon RX Vega running much faster. Instead, undervolting can achieve far better results.
First and foremost, the use of a suitable utility like OverdriveNTool works wonders. As always, though, your results will also depend on the quality of your GPU. We can't generalize; you'll have to compare your improvements to ours.
Temperatures & Frequencies
We’re using the GPU temperature value exclusively because that's what our test sample’s telemetry reports. Of course, the hot-spot temperature is a lot higher. Why? You can read all about this in Does Undervolting Improve Radeon RX Vega 64's Efficiency? On Asus' ROG Strix Radeon RX Vega 64, those readings are up to 15°C higher.
The following table shows a comparison of start and end values for temperatures and GPU (boost) frequencies. Just keep in mind that these clock rates can be considerably higher in games with significantly lower loads. For example, Wolfenstein 2's indoor environments might push the card to nearly 1.6 GHz, only to knock it way down once you step outside.
|Initial Value||Final Value|
|Open Test Bench|
|GPU Clock Rate||1408 MHz||1360 MHz|
|GPU Clock Rate||1403 MHz||1358 MHz|
|Air Temperature In Case||24°C||49°C|
Temperature vs. Frequency
To better illustrate our findings, we plotted temperatures and frequencies during our sample's 15-minute warm-up phase. It's particularly interesting that there's such a small thermal difference between open and closed cases.
Frequencies in the gaming loop are about 100 MHz higher than what we measured from AMD's reference card. This average increase of ~10% only improves gaming performance by 5-7%, which isn't particularly impressive.
The results of our stress test look similar:
IR Image Analysis Of The Board's Back
To round out this section, we take a look at board temperatures across several different load levels. To keep the test setup as practical as possible, we removed the backplate for IR measurements (since it doesn't help with cooling anyway) and reattached the frame using our own special nuts. Comparative before/after tests show no difference in temperature or cooling performance.
It's easy to see that the card is already at its limits during our The Witcher 3 benchmark. As long as you operate the ROG Strix Radeon RX Vega 64 8GB OC Edition vertically on an open test bench, all values seem to be acceptable. But this may change quickly if the card is installed in a closed case instead. Interestingly, VRM2, where you find the two separately located voltage converters, is significantly cooler.
Once we popped Asus' board into a closed case, we measured up to 4°C-higher temperatures at the voltage converters. It seems as though the fans only respond to GPU temperature, unfortunately. A slightly higher rotational speed or passive cooling on the board's back side would have helped mitigate such large hot-spots.
The stress test reflects slightly lower power consumption than our gaming benchmark. However, certain components (like the voltage converters) still get a little warmer. This is ultimately the result of a more constant load, which can be difficult to keep up with.
In a closed case, temperatures rise by three to four degrees Celsius.
With a PCB that's so nice and tidy on the back, Asus shouldn't have pinched pennies, leaving off a good thermal pad.
We don't like the heat build-up measured around this card's voltage regulation circuitry, especially since that thermal energy, over time, soaks into the GPU package, heating it from below. A bit of passive help from a thermal pad would have come highly recommended.
This is also the reason why we avoided cranking up the power limit or overclocking more aggressively. The card not only gets hot in a closed PC case, but it also makes a lot more noise.
MORE: Best Graphics Cards
MORE: All Graphics Content